FORECASTERS' FORUM Elevated Convection And

1280 WEATHER AND FORECASTING VOLUME 23

FORECASTERS’ FORUM

Elevated Convection and Castellanus: Ambiguities, Significance, and Questions

STEPHEN F. CORFIDI NOAA/NWS/NCEP/Storm Prediction Center, Norman, Oklahoma

SARAH J. CORFIDI NOAA/NWS/NCEP/Storm Prediction Center, and Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma

DAVID M. SCHULTZ* Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, and NOAA/National Severe Storms Laboratory, Norman, Oklahoma

(Manuscript received 23 January 2008, in final form 26 April 2008)

ABSTRACT

The term elevated convection is used to describe convection where the constituent air parcels originate from a layer above the planetary boundary layer. Because elevated convection can produce severe hail, damaging surface wind, and excessive rainfall in places well removed from strong surface-based instability, situations with elevated storms can be challenging for forecasters. Furthermore, determining the source of air parcels in a given convective cloud using a proximity sounding to ascertain whether the cloud is elevated or surface based would appear to be trivial. In practice, however, this is often not the case. Compounding the challenges in understanding elevated convection is that some meteorologists refer to a cloud formation known as castellanus synonymously as a form of elevated convection. Two different definitions of castellanus exist in the literature—one is morphologically based (cloud formations that develop turreted or cumuliform shapes on their upper surfaces) and the other is physically based (inferring the turrets result from the release of conditional instability). The terms elevated convection and castellanus are not synony- mous, because castellanus can arise from surface-based convection and elevated convection exists that does not feature castellanus cloud formations. Therefore, the purpose of this paper is to clarify the definitions of elevated convection and castellanus, fostering a better understanding of the relevant physical processes. Specifically, the present paper advocates the physically based definition of castellanus and recommends eliminating the synonymity between the terms castellanus and elevated convection.

1. Introduction within the convection lies above the planetary boundary layer (PBL). Specifically, elevated convection oc- The term elevated convection denotes convective curs above any near-surface stable layer (e.g., nocturnal clouds, storms, or both where the origin of air parcels inversion) or a sloping frontal surface (such as a warm or stationary front) where the instability is above the surface. Although the term elevated convection has * Current affiliation: Division of Atmospheric Sciences and Geophysics, Department of Physics, University of Helsinki, and been used more widely in recent years, the concept of Finnish Meteorological Institute, Helsinki, Finland. convection (particularly, deep moist convection) not based in the PBL has been in existence for many de- cades (e.g., Berry et al. 1945, pp. 714 and 816). Deep Corresponding author address: Stephen F. Corfidi, Storm Pre- diction Center, 120 David L. Boren Blvd., Ste. 2300, Norman, OK elevated convection, generally in the form of thunder- 73072. storms, can produce excessive rainfall, hail, and occa- E-mail: [email protected] sionally damaging surface winds and tornadoes in areas

DOI: 10.1175/2008WAF2222118.1

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

WAF2222118 DECEMBER 2008 FORECASTERS’ FORUM 1281 well removed from strong surface-based instability storms whose most unstable parcel is located at the (e.g., Branick et al. 1988; Schmidt and Cotton 1989; surface. But the fact that two otherwise identical “sur- Colman 1990a,b; Neiman et al. 1993; Grant 1995; Ber- face based” storms might have different levels of most nardet and Cotton 1998; Moore et al. 1998, 2003; Ba- unstable inflow at the very least calls into question the nacos and Schultz 2005; Goss et al. 2006; Colby and widely accepted notion that a simple dichotomy exists Walker 2007; Horgan et al. 2007), as well as lightning- between surface-based and elevated storms. initiated wildfires in the western United States (Tardy Efforts to determine storm inflow layers are moti- 2007). vated by the fact that the depth and location of a con- Determining the source region of buoyant parcels vective cloud’s inflow can affect its subsequent evolu- with the aid of an appropriate proximity sounding is not tion and tendency to produce severe weather. For ex- necessarily trivial. Specifically, growing cumulus clouds ample, Horgan et al. (2007) showed that elevated and thunderstorms based in the boundary layer rou- convection tends to be associated with a reduced like- tinely ingest parcels from above the boundary layer lihood of producing significant severe winds and torna- (e.g., Stull 1988, 559–561; Emanuel 1994, 200–204; and does. Specifically, of all severe-storm reports associated references therein). A modeling study by Fovell (2005) with the 129 elevated severe-storm cases in Horgan et suggests that deep convective clouds that form in the al. (2007), 9% of all hail reports, 3% of all wind reports, vicinity of sea-breeze circulations are composed of air and 10% of all tornado reports were significant severe that initially originates above the PBL. On the other reports, as defined by Hales (1988). These numbers hand, air from near-surface stable layers can be incor- compare to the 9.8% of hail, 15.8% of wind, and 18.3% porated into the updrafts of developing storms as long of tornadoes that are significant reports from all storms as the resulting parcels become positively buoyant. In during 1970–2004 (G. Carbin 2007, personal communi- addition, rotating updrafts in supercell storms induce cation). nonhydrostatic vertical pressure gradients that tap non- When, or even if, a cloud or storm transitions from buoyant boundary layer parcels (e.g., Marwitz 1973; being surface based to elevated, or vice versa, is often a Browning and Foote 1976; Weisman and Rotunno difficult forecasting challenge. For example, opera- 2000). To define surface-based convection as convec- tional experience and indirect evidence via visual ob- tion in which the air involved is derived mainly from the servation suggest that thunderstorms associated with PBL begs the question, “what is mainly?” deeply mixed boundary layer environments over the Thompson et al. (2007) provide one approach to an- Rocky Mountains and adjacent high plains of the swering this question by considering the effective inflow United States frequently become elevated as they move layer of a storm. Using proximity soundings to delimit east across the lower plains. Even without strong con- the vertical range of parcels meeting selected convec- vective inhibition, the cooler, but more moist, air from tive instability and inhibition criteria, the authors iden- the PBL over the lower terrain does not appear to be tify the layer that likely serves as the primary source for ingested into such storms, thereby reducing the likeli- a storm’s updrafts. This layer is then used to compute hood for tornadoes (e.g., Horgan et al. 2007). On the improved estimates of the magnitude of the environ- other hand, initially elevated storms sometimes clearly mental shear and storm-relative helicity (Davies-Jones do become surface based upon encountering regions et al. 1990) associated with an elevated storm. Critical with moister boundary layers. Drawing upon this discussion of the Thompson et al. technique for identi- moister air, the circulations of such storms appear to fying effective inflow layers is beyond the scope of this develop downward, often displaying a concomitant in- paper. Their scheme is, nevertheless, a first attempt to crease in strength and severity (e.g., Rockwood and better quantify the level of potential severity posed by Maddox 1988, 63–65). elevated storms. Further, Thompson et al. correctly Compounding the challenges in understanding el- note (p. 108) that the most buoyant parcel in a storm’s evated convection is that some meteorologists refer to a inflow layer often exists well above the surface, even type of cloud formation known as castellanus synony- with storms whose inflow layers include the surface mously as a form of elevated convection. Castellanus (e.g., their Fig. 6). Storms of this nature occur fre- (meaning “castle shaped”), in its most common usage, quently in the moist, marginally unstable environments is a patchy or streaky cloud formation with turrets (Fig. common to severe weather events over the southern 1), although castellanus can also be used to describe the and eastern United States. It is not clear if or how entire cloud containing such turrets or even a whole storms with elevated, most unstable parcels might differ field of such clouds. These cloud formations typically morphologically and behaviorally from surface-based represent comparatively benign convection with rela-

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1282 WEATHER AND FORECASTING VOLUME 23

FIG. 1. Castellanus (arranged in lines arising from a common base, lower left) and floccus (tufted, cumuliform puffs with ragged bases, top center through lower right) near Valentine, NE, about 1400 central daylight time (CDT; CDT ϭ UTC Ϫ 5 h) 28 May 1988, looking south.

tively weak updrafts and limited vertical extent. The questions requiring answers from the research and fore- structure and evolution of castellanus provide insight casting communities. into the relevant physical processes responsible for such cloud forms, as we discuss later in this paper. 2. Castellanus Given the challenges in understanding the origins of air associated with elevated convection, the physical a. Morphologically based definition processes involved, and the inconsistent terminology, Clouds have been classified since the late nineteenth the present paper addresses elevated convection and century using a scheme derived from that introduced by castellanus. We hope to not only clarify some aspects the English pharmacist Luke Howard in 1803 (Ham- that we feel heretofore have been neglected, but also blyn 2001). This system, based primarily on the shape pose questions and encourage additional discussion. In and appearance of clouds as seen from the ground, was particular, we would like to increase awareness of the adopted in modified form by the editors of the Inter- genesis and evolution of elevated convection so that national Cloud Atlas (hereafter the Atlas) in the early both understanding and forecasts may be improved. 1900s (World Meteorological Organization 1956). Ap- Section 2 of this article provides a discussion of castel- plication of the scheme facilitated the use of ground- lanus and two different definitions of castellanus be- based visual cloud observations in synoptic-scale me- cause castellanus is a familiar, visual vehicle by which teorological analysis, especially before the availability many meteorologists are introduced to the concept of of geostationary satellite data in the 1970s and the elevated convection. Examples of castellanus and el- implementation of the Automated Surface Observing evated convective clouds as seen from the ground are System (ASOS) network in the 1990s. presented in section 3 to illustrate some of the many The Atlas identifies 10 basic cloud types (genera) that forms that exist, and to show that the division between are separated into low, middle, and high categories elevated and surface-based convection often is indis- based on their commonly observed heights above the tinct. Why castellanus and elevated convection are im- ground (cirrus, cirrostratus, cirrocumulus, altostratus, portant to forecasting is discussed in section 4, includ- altocumulus, nimbostratus, stratus, stratocumulus, cu- ing the role that castellanus may play in the develop- mulus, and cumulonimbus). Given the emphasis in the ment of surface-based convection and the transition Atlas on cloud shapes rather than the physical processes from elevated convection to surface-based forms. Fi- involved in their formation, that specific nomenclature nally, section 5 concludes the paper, asking further for elevated convection does not exist on the Atlas is

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 1 live 4/C DECEMBER 2008 FORECASTERS’ FORUM 1283 not surprising. However, the Atlas (Vol. 1, p. 12) does Despite providing a more physical explanation for recommend the term castellanus to designate patches castellanus than the Atlas, Scorer’s definition has not or layers of cloud at any level that assume turreted or received widespread acceptance. Castellanus is still cumuliform parts on their upper surfaces, particularly widely considered to be strictly a midlevel cloud for- when the diameter of the towers is small relative to mation with turrets, more or less as described in the their heights. The turrets are connected by a common Atlas. This limited view, unfortunately, perpetuates the base and often are arranged in lines. This morphologi- notion of castellanus as a small subset of cloud forma- cal definition is consistent with that in the Glossary of tions set apart from other forms of moist convection. In Meteorology (Glickman 2000, p. 118). A typical contrast, Scorer’s broader approach recognizes that midlevel castellanus formation is shown in Fig. 1. In castellanus is part of a continuum that ranges from the places, the turrets are ragged and the cloud bases have decidedly PBL-based forms of cumulus described by dissolved entirely, presumably the result of updraft di- Stull (1985) to various elevated types such as altocumu- lution by entrainment of dry air. The Atlas (p. 12) refers lus castellanus and convective cirrus (cirrus uncinus). to castellanus of this type as floccus. If we allow that the Indeed, the Glossary of Meteorology (Glickman 2000, development of turrets on a stratiform cloud reflects, in p. 119), echoing the Atlas, implies such a physically part, the release of conditional instability (i.e., the tur- based description of castellanus by stating, “When al- rets result from the presence of condensation) at a level tocumulus castellanus and stratocumulus castellanus at- above the surface [e.g., “high-based convective clouds,” tain a considerable vertical development, they become as described by Houze (1993, 189–191)], then at least cumulus congestus and often develop into cumulonim- some clouds that fit the Atlas’s description of castella- bus.” Hereafter in this article, we adopt Scorer’s defi- nus are necessarily elevated—even though the word el- nition of castellanus because it emphasizes the specific evated is not part of the Atlas’s definition. physical process that distinguishes castellanus from Casual observation of the sky reveals that turreted other forms of convective clouds. Such an approach clouds can occur at all levels in the troposphere. In acknowledges that castellanus is not limited to any par- practice, largely because of long-standing requirements ticular subset of genera, but rather represents an im- by official meteorological codes that observed clouds portant process that can occur with varying frequency be classified into 1 of the 10 genera, and because earlier and with minor variation at all levels of the troposphere. editions of the Atlas specifically associated castellanus with the genera altocumulus, castellanus has come to be 3. Examples of elevated convection and viewed almost exclusively as a formation typically as- castellanus sociated with midlevel clouds of the genus altocumulus. In the United States, association of castellanus with In this section, several forms of elevated convection altocumulus has been furthered by widespread adop- and castellanus are presented. The examples are not tion of the aviation acronym for altocumulus castella- meant to be all inclusive; instead, through photographs nus (ACCAS) in surface airway observations and op- we emphasize how these formations appear visually, erational weather discussions. and how they reflect the environmental processes responsible for their development. The presentation be- b. Physically based definition gins with two classic forms of castellanus—forms that In contrast to the morphologically based definition in most likely come to mind when one hears the word the Atlas, Scorer (1972, p. 31) offers a physically based castellanus. We start with castellanus because the for- definition of castellanus: any cumuliform cloud forma- mation has long been recognized as a distinctive cloud tion that owes its buoyancy to the occurrence of con- form, whereas elevated convection is a term without densation, rather than to the presence of thermals clear visual representation. reaching the level of free convection (LFC). The con- Figure 1 shows patchy midlevel castellanus of the densation can result from any number of processes, in- type that frequently accompanies regions of ascent cluding uplift within orographically induced waves, as- ahead of midtropospheric disturbances in the wester- cent of potentially unstable air ahead of midtropo- lies. Such cloud formations are especially common in spheric disturbances, and the saturation of layers elevated mixed layers extending downstream from ascending beneath widespread precipitating cloud high-level heat sources such as the Mexican plateau and decks such as those associated with upper lows. The Rocky Mountains. Castellanus is a visual manifestation examples in Scorer (1972, 31–37) illustrate that castel- of the release of conditional instability as a result of lanus can be quite common and can include formations large-scale isentropic ascent of shallow moist layers in not traditionally considered castellanus. the elevated mixed layer (e.g., Carlson and Ludlam

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1284 WEATHER AND FORECASTING VOLUME 23

FIG. 2. (a) Western edge of an extensive area of deep castellanus associated with weak 850–700-hPa warm advection on the leading edge of an elevated mixed layer near Norman, OK, 0800 CDT 13 Jul 2006, looking southwest. (b) Rawinsonde observation at 0700 CDT 13 Jul 2006 at Norman, OK. Wind speeds in knots. The sounding suggests that the clouds in Fig. 2a were based around 600 hPa. (c) Elevated thunderstorms forming above a slowly moving warm front near Omaha, NE, at about 1830 CDT 25 Jun 1994, looking south. (d) Rawinsonde observation at 1900 CDT 25 Jun 1994 at Omaha, NE. Wind speeds in knots. The sounding suggests that the storms in Fig. 3a were based around 775 hPa.

1968). The moist layers may be derived from the evapo- thunderstorms (e.g., Ley 1894; World Meteorological ration of ordinary PBL cumuli within the elevated Organization 1956; Ludlam and Scorer 1957, p. 39), and mixed layers at points upstream from the castellanus or statistical evidence exists to support this (e.g., Brooks could reflect layers where PBL cumuli spread out (an- 1951). viled) beneath weak inversions in the elevated mixed If the elevated moist layer in a region of isentropic layers (Ludlam 1980, p. 246). The castellanus usually ascent is comparatively deep, castellanus may develop first appear in what may be the crests of low-amplitude sufficient depth to produce showers. Figure 2a provides gravity waves set up by the underlying topography (e.g., an early morning view of the western edge of an exten- Ludlam and Scorer 1957, p. 39; Scorer 1972, p. 22). sive area of deep castellanus associated with an area of Initially laminar, the clouds subsequently break into weak 850–700-hPa warm advection at the leading edge cumuliform turrets aligned with the mean shear at their of an elevated mixed layer over the southern plains. A level as latent instability is released through condensa- rain shower is present just beyond the left side of the tion. In many parts of the world, castellanus formations photo. The smooth, laminar cloud bases (Fig. 2a, top on clouds have long been considered forerunners of left) are especially characteristic of deep and/or wide-

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 2a c live 4/C DECEMBER 2008 FORECASTERS’ FORUM 1285 spread areas of castellanus and reflect the isentropic 205–211), the only real difference being that nimbo- ascent initially responsible for their development. Simi- stratus-generating cells are embedded in widespread lar castellanus formations with extensive laminar bases light precipitation, whereas cirrus cells are surrounded appear frequently above thunderstorm outflow bound- by clear sky. Cirrus-generating cells appear as small, aries. The proximity sounding in Fig. 2b suggests that white puffs above the fallstreaks in the top-left part of the clouds were based around 600 hPa. Fig. 3a; the clouds were between 375 and 325 hPa ac- Areas near warm and stationary fronts are frequent cording to the sounding in Fig. 3b. If one accepts Scor- sites of deep elevated convection. As previously noted, er’s definition of castellanus, the generating cells of cir- such convection is of considerable interest given its po- rus uncinus are, in fact, an aberrant form of castellanus. tential to produce severe weather, even well within the Long-lasting, regenerative bands of castellanus cold air (e.g., Neiman et al. 1993; Grant 1995; Trapp et sometimes form on the leading edge of forward- al. 2001; Horgan et al. 2007). Figure 2c shows elevated propagating midlatitude mesoscale convective systems thunderstorms forming about 150 km north of a slowly (MCSs). Such formations are most apparent when the moving warm front. The convection is sprouting from cold pool and gust front have begun to outrun the main an extensive band of laminar clouds that appear to be convective core late in the MCS’s life cycle and the based at the same level (around 775 hPa, per radio- system has developed a significant component of front- sonde data in Fig. 2d) as the patchy wave clouds in the to-rear flow (Fig. 3c). Radar observations of MCSs with foreground. These storms and others farther south (in regenerative elevated convective bands similar to Fig. the direction of this view) continued to strengthen after 3c suggest that, as the cold pool deepens beneath the the photo was taken. Cloud-layer shear was in excess of clouds (in response to the continued advance of the 25 m sϪ1 (Fig. 2d). Supercell storms that evolved from deep convection responsible for the cold pool’s devel- this convection produced a tornado in northwest Mis- opment), the towers also deepen and subsequently souri, even though conventional surface data and visual merge with the parent MCS. In this manner, the cas- observations of laminar low- to midlevel clouds tellanus becomes an integral part of the convective sys- strongly suggested that the updraft bases were elevated tem. The cloud formation in Fig. 3c has a great hori- on the cool side of the front. zontal extent (the band continued in both directions When the environment is sufficiently cold and ice beyond the field of view) and a smooth base typical of nuclei are present, convective clouds with limited ver- elevated convection. The formation changed little over tical extent (e.g., altocumulus) can glaciate, producing time, and there was an absence of convective towers trails of ice crystals known as fallstreaks (e.g., Scorer ahead of it. Together, these observations suggest that 1972; Hobbs and Rangno 1985). Over time, fallstreaks the castellanus band reflects the presence of a broad, become aligned with the shear at their level, assuming slablike swath of forced ascent on the leading edge of a hooked shape when speed or directional shear is the MCS, as discussed by Bryan and Fritsch (2001). The marked (Fig. 3a). Because of the low saturation vapor distribution, intensity, orientation, and overall charac- pressure of ice in their environment, fallstreaks evapo- ter of the radar echoes of the convective system asso- rate slowly (compared to liquid or supercooled water ciated with the cloud band shown in Fig. 3c (not shown) droplets) and tend to persist over time. Distorted by the bear close resemblance to the example in Bryan and wind, the fallstreaks often assume a fibrous texture Fritsch (their Fig. 11). As Bryan and Fritsch (2001) when seen from the ground. noted, such areas of ascent often exhibit moist absolute Fallstreaks and their parent convective clouds are instability (i.e., persistent saturation in layers with lapse commonly referred to as cirrus uncinus (“hooked cir- rates greater than moist adiabatic). Indeed, a proximity rus”), even though such formations often originate sounding made just ahead of the convective system from altocumulus and may occur at altitudes lower than shows a moist, absolutely unstable layer between 750 those commonly associated with cirrus (at or above 6 and 650 hPa above a more stable layer at the surface (Fig. km AGL in the midlatitudes according to the Atlas). 3d). Thus, some castellanus formations may be mani- Although cirrus uncinus appear quiescent to casual ob- festations of the release of moist absolute instability.1 servation, such clouds actually are in a state of continu- ous redevelopment. The parent convective elements, known as generating cells (e.g., Heymsfield 1975; Lud- 1 Because soundings are nominally vertical, yet flow toward and lam 1980; Atlas 2001), constantly re-form as latent heat atop the leading edge of an MCS gust front is quasi-horizontal, the true meaning of soundings that contain moist absolutely unstable is released via condensation and freezing. Similar gen- layers is open to question. Our point here is that castellanus erating-cell–fallstreak couplets enhance precipitation in clouds are sometimes present when nearby soundings depict moist deep nimbostratus cloud systems (e.g., Houze 1993, absolute instability.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1286 WEATHER AND FORECASTING VOLUME 23

FIG. 3. (a) Hook-shaped fallstreaks trailing from elevated shallow convection over Norman, OK, at 1845 CDT 14 Sep 2006, looking north. (b) Rawinsonde observation at 1900 CDT 14 Sep 2006 at Norman, OK. Wind speeds in knots. The sounding suggests that the clouds in Fig. 3a were between 325 and 375 hPa. (c) Castellanus band above the gust front of an approaching MCS at Kansas City, MO, at 0730 CDT 16 Jun 1996; wide-angle view looking west-northwest. The sky above the castellanus is covered by anvil material spreading toward the observer ahead of the MCS. (d) Rawinsonde observation at 0700 CDT 16 Jun 1996 at Topeka, KS; at the time of the sounding, the observation site was located just east of the MCS shown in Fig. 3c. A moist, absolutely unstable layer is present between 750 and 650 hPa. Wind speeds in knots.

The examples of castellanus and elevated convection lanus can originate at any level, including the PBL. Two presented thus far suggest that the updrafts were not examples of PBL-based castellanus are shown in Figs. based in the boundary layer. Implicit in Scorer’s defi- 4a,b. This form of castellanus is most common in areas nition of castellanus, however, is the notion that castel- where surface-based updrafts tend to be weak (e.g.,

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 3a c live 4/C DECEMBER 2008 FORECASTERS’ FORUM 1287

FIG. 4. Two examples of castellanus towers rising from foundations of shallower PBL clouds over the tropical western Atlantic. [Taken near Barbuda on 16 Jan 2005 during the Rain in Cumulus over the Ocean (RICO) field campaign.] (c) PBL castellanus at sunset, forming in the crests of waves left moistened by ordinary diurnal boundary layer cumuli over Norman, OK, at 2044 CDT 1 Jul 2006, looking north. Rawinsonde data suggest that the clouds were based around 700 hPa. over oceanic regions in low latitudes). Feeble boundary level convergence; as a result, they are quickly over- layer convergence in such environments can form con- whelmed by entrainment of dry air and become spindly. vective clouds with limited vertical extent that subse- The narrow towers of such clouds contrast with the quently deepen through latent heat release. Clouds like broader outlines of true cumulus congestus, the sustain- those in Figs. 4a,b are not associated with strong low- ing parcels of which involve the depth of the PBL and

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 4a b c live 4/C 1288 WEATHER AND FORECASTING VOLUME 23 are supported by more persistent convergence. Some- outflow produced by an area of midlevel castellanus what analogously, so-called “turkey-tower” cumulus altered low-level convergence along a cold front. The congestus (e.g., Fig. 1.4 in Doswell 2001), wherein deep, cold front, oriented west–east and marked by the re- PBL-based convective towers rapidly grow but subse- gion of featureless low clouds over northern Wisconsin quently quickly dissolve from the base up, might be (Fig. 5a), was preceded by an area of the castellanus considered a hybrid form of castellanus by the Atlas’s over the central part of the state. West-northwesterly definition, in the sense that the towers are taller than midlevel flow mixed to the surface within the castella- they are wide. nus outflow (Figs. 5a and 5b). This weakened conver- Another variety of PBL-based castellanus is shown in gence along the front as southwesterly surface winds in Fig. 4c. Clouds of this type are occasionally observed the undisturbed warm sector became light and variable around sunset following a day of diurnal convection or west-northwesterly in the wake of the outflow (Figs. with relatively limited vertical extent. As with other 5b,c). The clouds also diminished heating in the pre- forms of castellanus, the turrets grow from patches of frontal warm sector. These factors appeared to be at cloud that develop in the crests of shallow orographic least partially responsible for the absence of deep con- lee waves. In this case, however, the waves exist at the vection along the front throughout most of the day top of the PBL, which has been left moistened by the (Figs. 5a and 5b). In contrast, surface-based thunder- evaporation of the previous afternoon’s cumulus (Lud- storms with hail formed during the afternoon at the lam and Scorer 1957). Formations like the one in Fig. 4c intersection of two castellanus-derived outflow bound- often are dismissed as being the dying remnants of or- aries in southeast Wisconsin (Fig. 5c; an animation of dinary diurnal cumulus and, prior to the most recent this imagery is available at the Journals Online Web edition of the complete Atlas (1956), were known as site at http://dx.doi.org/10.1175/2008WAF2222118.s1). stratocumulus vesperalis (“evening stratocumulus”). In that region, the outflow boundaries encountered un- Careful observation, however, reveals that the turrets modified warm, humid southwesterly flow in the pre- arise from recently formed patches of wave clouds; the frontal warm sector, where surface heating and conver- turrets, therefore, likely derive their buoyancy from gence were sufficient to initiate surface-based storms. condensation in the waves. Normally, such clouds are of Cases like this illustrate how the location and evolution little significance as they quickly dissipate upon entrain- of deep surface-based convection can be affected by the ing drier air from above the PBL. Operational experi- presence of castellanus clouds. ence, however, has shown that such clouds, particularly Another example of how elevated convection may over the central United States, may mark the initial affect subsequent surface-based storm development is stages of elevated nocturnal thunderstorms, forming on shown in Fig. 6. Figure 6a, made shortly after sunrise, the leading edge of a moisture gradient associated with shows an elongated field of castellanus with bases near a strengthening low-level jet stream. Such formations 700 mb (per regional soundings; not shown) over cen- therefore bolster the notion of castellanus as a harbin- tral and eastern Oklahoma. Cloud-layer winds were west- ger of thunderstorms. southwesterly at 20 kt (10 m sϪ1) on the poleward side PBL-based castellanus illustrate that the partition be- of a subtropical ridge. The location and spatial extent of tween purely elevated and purely surface-based forms diurnal thunderstorms that formed over northern and of convection is far from distinct. Such clouds, as with western Arkansas later in the day (Fig. 6b) bear close the other types presented earlier, also remind us that, in resemblance to the extrapolated outlines of the morn- failing to adopt a more precise classification scheme for ing castellanus field over Oklahoma (an animation of castellanus and elevated convection, we may overlook this imagery is available at the Journals Online Web valuable clues that such clouds provide about the state site at http://dx.doi.org/10.1175/2008WAF2222118.s2). of the atmosphere in their vicinity, and about convec- We speculate that the field of castellanus clouds tion initiation in general. marked a region of enhanced midlevel moisture that reduced entrainment of dry air as it moved downstream 4. Some practical aspects of elevated convection across developing PBL-based convection in Arkansas. and castellanus Some of the most challenging forecast situations in- Although elevated convective clouds with limited volving elevated convection are those in which the vertical extent often are relatively unimportant for con- clouds deepen, redevelop nearby, or both, ultimately vective storms forecasting, occasions occur when such becoming surface based. Cases of this type herein are clouds intimately are tied to the development of signif- referred to as transition events. Questions as to if, when, icant convective weather. The satellite sequence in Fig. and where a transition event will occur are forecast 5 illustrates a situation of this type in which convective challenges because the processes responsible for the

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC DECEMBER 2008 FORECASTERS’ FORUM 1289

FIG. 5. Visible satellite data and surface observations (English units) over WI and Lake Michigan at (a) 1215, (b) 1415, and (c) 1615 CDT 8 Sep 2006. Mottled clouds over the central parts of WI and Lake Michigan in (a) and (b) are castellanus based near 700 hPa (per area rawinsonde data). Winds at this level were west-northwest at 15 m sϪ1. “MKE” and white square in (b) denote location of Milwaukee. Pertinent features mentioned in the text are shown in (c), where ACCAS is used to denote altocumulus castellanus. Two castellanus-derived outflow boundaries are visible over southern WI and southern Lake Michigan in (c); they intersect near the developing thunderstorm (“T-storm”) west of Milwaukee. The outflow boundaries are identifiable via surface observations in (a) and (b), and become apparent in the satellite imagery in (b) and (c) as they serve as a focus for the development of afternoon boundary layer cumulus clouds.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1290 WEATHER AND FORECASTING VOLUME 23

FIG. 6. Visible data satellite data and surface observations (English units) over OK and AR at (a) 0715 and (b) 1515 CDT 14 Aug 2006. Area soundings suggest that the castellanus clouds in (a) were based near 700 hPa; winds at this level were west-southwest at 10 m sϪ1. transition are themselves difficult to forecast. For ex- In the transition event shown in Fig. 7, elevated thun- ample, the strength and areal extent of convective in- derstorms developed within bands of morning castella- hibition (i.e., thermodynamic profiles), the location and nus over central Nebraska near North Platte (Figs. 7a– depth of outflow boundaries, and spatial and temporal c). The castellanus formations were based near 550 hPa changes in mesoscale forcing for ascent all can affect (Fig. 7g). The clouds are distinguishable by their frail the likelihood for transition. and tattered appearance in the satellite imagery (Figs.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC DECEMBER 2008 FORECASTERS’ FORUM 1291

FIG.7.(a)–(c) Visible data Geostationary Operational Environmental Satellite (GOES) imagery over part of the central United States, and (d)–(f) base reflectivity radar data from Omaha, NE, on 13 Jul 2006, with conventional surface data (English units) overlay and ACCAS used to denote altocumulus castellanus. Times are (a) 0815, (b) 1015, (c) 1215, (d) 1305, (e) 1505, and (f) 1705 CDT. LBF, OMA, and white squares in (b) denote locations of North Platte and Omaha. Rawinsonde observations at 0700 CDT 13 Jul 2006, at (g) North Platte, NE, and (h) Omaha, NE, and at 1300 CDT 13 Jul 2006 at (i) Omaha, NE. Wind speeds in knots. Elevated thunderstorms forming from midlevel convection in (c) became surface based during the early afternoon [shortly after the time of (d)] as surface heating eliminated inhibition present earlier in the day [cf. soundings (h) and (i)]. Surface-based storms forming on outflow boundary in lower-left part of (f) subsequently evolved into a new MCS.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1292 WEATHER AND FORECASTING VOLUME 23

FIG.7.(Continued)

7a–c), and by their comparatively fast motion relative over southeast Nebraska as the boundaries moved into to the surface winds (an animated loop of the imagery the region where 1) convective inhibition was weaker in Figs. 7a–c is available at the Journals Online Web site (cf. Figs. 7g and 7h) and 2) diurnal heating further less- at http://dx.doi.org/10.1175/2008WAF2222118.s3). This ened inhibition (cf. Figs. 7h and 7i). Rapid increases in convection produced outflow boundaries that later radar reflectivity, cross-sectional radar data (not shown), served as a focus for surface-based storm development and changes in the speed and direction of storm motion

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 7d live 4/C DECEMBER 2008 FORECASTERS’ FORUM 1293

FIG.7.(Continued)

(an animated loop of radar reflectivity over eastern Ne- heated environment near Omaha during the early af- braska is available at the Journals Online Web site at ternoon (Fig. 7d). Outflow from these more vigorous http://dx.doi.org/10.1175/2008WAF2222118.s4) all sug- storms produced strong convective outflows (Fig. 7e) gest that some of the elevated storms in Fig. 7c them- that subsequently fostered the development of addi- selves became surface based as they moved into the tional surace-based storms in southeast Nebraska later

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 7e f live 4/C 1294 WEATHER AND FORECASTING VOLUME 23

FIG.7.(Continued)

in the day (Fig. 7f). These storms, in turn, merged with tem left a path of destruction that included 50 m sϪ1 wind other developing surface-based cells to form a larger- gusts and grapefruit-sized hail in Lahoma, Oklahoma scale, forward-propagating MCS that moved south into (Janish et al. 1996). Another derecho that evolved from parts of Missouri and Kansas; the MCS produced dam- convection believed to have been at least partly elevated aging wind the following evening. occurred over northern Kansas and Missouri on 2 June One of the more dramatic transition events over the 1982 (Rockwood and Maddox 1988). In both of these United States in recent years occurred on 17 August 1994, cases, rapid spatial and temporal changes in boundary when an area of castellanus in southern Kansas evolved layer instability and inhibition were observed in the into an intense derecho. Supercells in the convective sys- areas where the convection became surface based.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC DECEMBER 2008 FORECASTERS’ FORUM 1295

The Lahoma and Kansas–Missouri events attained moved into northwest Oklahoma. Cases such as this maximum intensity shortly after the elevated convec- illustrate that the line between elevated and surface- tion moved or developed into a region experiencing base convection often is indistinct. Such systems also strong low-level destabilization (mainly in the form of raise questions as to if, how, and when gradually low- moisture advection). In contrast, Coniglio et al. (2007) ering, elevated convection will produce damaging sur- present an example of an elevated severe-wind- face wind. producing MCS that weakened as it moved from the Although forecasting the behavior of convective sys- cool side toward the warm side of an Oklahoma cold tems like the one just described is problematic, storms front. Part of the radar evolution of this case is shown that remain elevated throughout their lifetimes, yet in Fig. 8. The front is oriented roughly parallel to the produce damaging surface winds, are also very chal- southern edge of the lifted-index maximum that ex- lenging. An elevated supercell that produced significant tends from the northern Texas Panhandle into north- wind damage in northeast Kansas and northwest Mis- central Oklahoma in Figs. 8a–d. The nocturnal convec- souri on the morning of 12 March 2006 provides an tive system evolved from an elevated supercell in south- example (Goss et al. 2006). The storm left a swath of east Colorado (Figs. 8a and 8b) to a bow echo that surface temperature rises in its wake. Proximity sound- produced gusts in excess of 30 m sϪ1 across southwest ings suggest that the warming may have resulted from Kansas and northwest Oklahoma (Figs. 8b–d). The for- penetration of storm-induced saturated downdrafts ward-propagating system began to weaken as it en- through a shallow, but very cool, boundary layer (pre- countered higher surface-based instability close to the storm surface temperatures were around 8°C). A simi- surface boundary (Figs. 8e and 8f; an animated loop of lar event occurred over southeast Kansas on 8 April reflectivity is available at the Journals Online Web site 2008, as this paper was in the final stages of prepara- at http://dx.doi.org/10.1175/2008WAF2222118.s5). tion. Both cases involved strong low-level warm advec- Based on this, one might conclude that kinematic dif- tion, with a shallow (approximately 0.5 km deep) layer ferences along the MCS’s path evidently were respon- of cold air surmounted by a deep elevated mixed layer sible for the system’s decline. For example, wind pro- that extended from near 850 to 500 hPa (not shown). files were indeed less favorable for renewed cell devel- The storms appeared to be based within 50-hPa-deep opment on the downwind (southeast) side of the system saturated layers based near 800 hPa (i.e., within the cold pool on the cool side of the front than they were on elevated mixed layers). Events like these are particu- the warm side. On the cool side, easterly low-level larly vexing as elevated supercells in similar environ- winds were surmounted by west-to-northwest flow at ments are routinely observed without intense surface mid- and upper levels. Such a setup can foster conver- gusts. gence and new cell development on the downwind side Elevated thunderstorms that occur in environments of the system cold pool (Corfidi 2003). of strong warm advection near fronts are sometimes Although kinematic factors may have been involved, arranged regularly spaced in a line oriented roughly evidence also suggests that the MCS likely weakened parallel to the isentropic surfaces in their vicinity (Fig. because it moved beyond an axis of maximum elevated 9a). In other cases storms appear in short bands ori- instability as it neared the surface front. This instability ented roughly perpendicular to the isentropes (Fig. 9b). axis was associated, in part, with a plume of steep The storms in Fig. 9a, some of which are supercells, are midlevel lapse rates that extended east-northeast from evenly spaced at intervals of approximately 11⁄2 storm northern New Mexico (Coniglio et al. 2007, their Figs. diameters. In Fig. 9b, the spacing of the elongated 15 and 16). The lifted index data in Figs. 8a–f show that storms varies from about 1 storm width along the Mis- surface-based instability was indeed higher near the souri–Illinois border to several storm widths in south- front than it was farther north. Evidently, this differ- east Illinois. Other degrees of separation also have ence in lifted indices across the front was not sufficient been observed, and storm spacing sometimes varies to offset the more hostile midlevel thermodynamic en- over time. The factors governing the regular arrange- vironment and reduced magnitude of cold-pool-relative ment of elevated convective cells are not known, and flow that existed near and south of the front. the responsible processes likely are not the same in all The MCS just described involved system propagation cases. Storm spacing does not appear to be directly re- (i.e., new cell development) via surface or near-surface lated to the spacing of horizontal roll circulations in the convergence at the leading edge of the system cold pool upstream warm sector, nor does it appear to be a func- during much of its existence. Yet, surface (Fig. 8), pro- tion of a readily observable characteristic of the storms filer, and sounding data (not shown) indicate that the themselves. Depending upon the motion of the storms system clearly was elevated, at least as the system relative to their arrangement, some locations may ex-

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1296 WEATHER AND FORECASTING VOLUME 23 perience repeat episodes of hail, wind, and rain as in- which castellanus-type formations might affect the evo- dividual cells pass by. To our knowledge, documenta- lution of sensible weather. tion of regularly arranged elevated storms has not heretofore appeared in the literature. (Animated loops of 5. Concluding remarks radar reflectivity for the cases presented in Fig. 9 are This paper has presented a discussion of elevated available at the Journals Online Web site at http://dx. convection and castellanus. Our aim has been to en- doi.org/10.1175/2008WAF2222118.s6 and http://dx.doi. courage more critical thinking about these cloud for- org/10.1175/2008WAF2222118.s7.) mations and to demonstrate that our questions are not The first case in this section (Fig. 5) illustrated a way purely academic—they have real-world forecasting ap- in which castellanus can affect surface-based storm de- plication. This section briefly summarizes what has thus velopment, and we noted in section 3 that the growth far been presented and provides a few additional and movement of some MCSs involve castellanus for- thoughts to foster further discussion. mations above the system outflow boundary. In a re- Routinely in the scientific literature and in opera- cent numerical investigation, Fovell et al. (2006) pre- tional forecast discussions, convection is classified de- sented evidence that discrete propagation of MCSs finitively as either surface based or elevated. Given sometimes involves the development of castellanus- what has been presented in this paper, such statements type formations in areas well removed from the gust appear to reflect greater confidence than the current front. These clouds form when conditions are favorable understanding of convective processes supports. Rather for the maintenance of internal gravity waves ahead of than furthering a binary point of view of convective cloud and storm behavior, we argue for a continuum of the convective system. In their simulation, a shallow convection, ranging from purely surface-based types to low- to midlevel cloud deck is first produced by low- purely elevated forms. A similar continuum, we sug- frequency gravity waves excited by the main convective gest, exists between those forms driven predominantly cluster of an MCS. This cloud layer then serves as the by latent heat release and those associated with ther- seat for new cell formation ahead of that cluster. For- mals rising through the LFC. Between these extremes, mation of the new cells occurs in tandem with passing an individual convective cloud or storm may exhibit high-frequency gravity waves associated with individual varying “degrees of elevation,”“castellanusness,” or convective bursts in the storm system. Some of the new both. Over time a given cloud or cloud system also may cells augment the original MCS by merging with it; become more or less elevated or more or less castella- other cells, meanwhile, appear to have a detrimental nus-like in response to changing environmental condi- effect on the original storm system. Those cells that tions (as illustrated, e.g., by the case in Fig. 7). have a negative impact ultimately supplant the original A range of convective clouds sometimes is apparent updraft cluster, resulting in discrete propagation (i.e., associated with midlatitude upper-level disturbances. redevelopment) of the convective system in the down- Figure 10, for example, is a visible data satellite image wind direction. Details regarding the origin and evolu- showing part of the cloud system of a weak trough in tion of updraft development in the shallow cloud layer, the westerlies over the south-central United States. The and how the convective cells relate to some of the cloud disturbance is typical of those in relatively dry regimes formations discussed in this paper are unclear. How- where obscuration by low stratiform clouds is minimal. ever, the new convective towers do not arise from con- The axis of the trough, denoted by a dashed white line vergence or heating at the surface, and system propa- in Fig. 10, is located just east of the New Mexico–Texas gation is not due to uplift along the leading edge of the border and is marked by deep surface-based convection cold pool. Situations of the type modeled by Fovell et (thunderstorms). This convection grades into increas- al. (2006) most often occur at night when the lower- ingly elevated forms with eastward extent over Kansas tropospheric relative humidity is greatest. These situa- and Oklahoma. The leading edge of the trough- tions also require a forward-tilted anvil to serve as a associated cloud system is composed of shallow, highly duct for the high-frequency gravity waves. Coinciden- elevated and glaciated convection; these clouds were tally, these conditions often are satisfied during the seen from the ground in Fig. 3a. Similar gradations in later stages of MCSs when, as discussed in section 3, a convection are also commonly seen in the cloud sys- link may exist between castellanus and the presence of tems of stronger midlatitude disturbances. Recognizing moist absolute instability. Although our intent is not to the indistinct border that exists between surface-based imply that moist absolute instability is necessarily in- and elevated forms of convection is key toward achiev- volved in the form of discrete propagation modeled by ing a better understanding of convective phenomena in Fovell et al., their work does suggest yet another way in general.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC DECEMBER 2008 FORECASTERS’ FORUM 1297

FIG. 8. Composite reflectivity radar sequence over southeast CO, southwest KS, and northwest OK at (a) 2343, (b) 0043, (c) 0143, (d) 0243, (e) 0343, and (f) 0443 CDT 30 Jun–1 Jul 2005, showing an elevated, forward-propagating MCS north of a west-southwest–east-northeast-oriented front. The MCS weakened as it moved into the region of maximum surface-based instability (most negative lifted index values) near the warm side of the front. Conventional surface data overlay (English units) with surface-based lifted index depicted by brown (values greater than and equal to zero) and green (values less than zero) contours (°C).

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 8a b live 4/C 1298 WEATHER AND FORECASTING VOLUME 23

FIG.8.(Continued)

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 8c d live 4/C DECEMBER 2008 FORECASTERS’ FORUM 1299

FIG.8.(Continued)

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 8e f live 4/C 1300 WEATHER AND FORECASTING VOLUME 23

FIG. 9. (a) Composite reflectivity data showing a line of evenly spaced elevated thunderstorms, some supercells, over southern WI at 0600 CDT 4 Oct 2006. Conventional surface data locate west–east- oriented cold front along radar fineline over northern IL. Surface-based warm sector thunderstorms are present over northern IL and northwest IN. (b) Composite reflectivity data showing elevated showers and thunderstorms in short bands over southern IL at 1800 CDT 26 Mar 2008. Conventional surface data show west–east-oriented stationary front extending from southern MO into western KY.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC

Fig 9a b live 4/C DECEMBER 2008 FORECASTERS’ FORUM 1301

FIG. 10. Visible data satellite image over part of the south-central United States on the occasion shown in Fig. 4 (1801 CDT 14 Sep 2006). Illustration shows the cloud system of a weak upper-level trough in the westerlies, the axis of which (dashed white line) was in west TX at the time of the image. Surface- based convection (including thunderstorms) over eastern NM and western parts of the Texas Panhandle grades into increasingly elevated forms of convection with its eastward extent over the central and eastern parts of OK and KS. Rawinsonde data indicate that the “midlevel” convection was based around 500 hPa, with tops around 300 hPa, while the “highly elevated” convection existed between 350 and 300 hPa.

Ignoring the shades of gray that exist in the natural castellanus are associated with elevated convection. We world is one hallmark of bad science; employing mul- further argue that different forms of castellanus exist, tiple definitions for the same term is another. The ex- ranging from traditional displays (e.g., Fig. 1) to those istence of more than one definition for castellanus is that are more aberrant (e.g., Figs. 2a, 3a, and 4). not in the best interest of science. Restricting castella- The discussion in the preceding paragraphs leads us nus to its traditional morphological definition certainly to propose the following classification scheme for con- has appeal based in familiarity. However, cloud defini- vection in the form of a Venn diagram (Fig. 11). Con- tions based on the primary physical processes involved vection can be either PBL-based or elevated, with gra- in their formation and evolution, rather than simply their appearance, seemingly best serve forecasters and researchers alike. Recognizing the improved understanding of the physical processes involved in cloud formation since the last complete International Cloud At- las in 1956, and given that automated observation systems and geostationary satellite data have more or less supplanted human surface cloud observations in the last quarter century, the time has come to move toward a more physically based system of cloud classification, as argued nearly a half century ago by Scorer (1963). Short of advocating a completely new classification scheme based on physical processes, we suggest restricting the use of the term castellanus to turreted FIG. 11. Venn diagram of the relationship between elevated cloud forms owing their buoyancy chiefly to the release convection, PBL-based convection, and castellanus. The shading of latent heat. We argue that castellanus can occur at represents a gradation between elevated and PBL-based convec- any level in the troposphere and that not all forms of tion.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC 1302 WEATHER AND FORECASTING VOLUME 23 dations in between. Within this spectrum lie castellanus homa Cooperative Agreement NA17RJ1227, U.S. De- cloud formations. Castellanus can be associated with partment of Commerce. either PBL-based convection (e.g., Fig. 4) or elevated convection (e.g., Figs. 1 and 2a). At the same time, not REFERENCES all elevated convection (e.g., the deep moist convection Atlas, D., 2001: Commentary and analysis: Fallstreaks and their in Fig. 2c) is associated with castellanus, and not all parent generators. Bull. Amer. Meteor. Soc., 82, 477–480. castellanus (e.g., PBL-based convection featuring cas- Banacos, P. C., and D. M. Schultz, 2005: The use of moisture flux tellanus; Fig. 4) is a form of elevated convection. convergence in forecasting convective initiation: Historical Section 4 of this paper presented several cases to and operational perspectives. Wea. Forecasting, 20, 351–366. illustrate the practical significance of castellanus and Bernardet, L. R., and W. R. Cotton, 1998: Multiscale evolution of a derecho-producing mesoscale convective system. Mon. elevated convection. These cases provide inspiration Wea. Rev., 126, 2991–3015. for many unanswered questions. For example, 1) Why Berry, F. A., Jr., E. Bollay, and N. R. Beers, Eds., 1945: Hand- is castellanus frequently banded, what determines the book of Meteorology. McGraw–Hill, 1068 pp. spacing of the bands, and what factors influence the Branick, M. L., F. Vitale, C.-C. Lai, and L. F. Bosart, 1988: The diameter of individual convective towers within them? synoptic and subsynoptic structure of a long-lived severe convective system. Mon. Wea. Rev., 116, 1335–1370. 2) Why do some elevated thunderstorms produce se- Brooks, C. F., 1951: The use of clouds in forecasting. Compen- vere surface winds whereas most do not? 3) What con- dium of Meteorology, T. F. Malone, Ed., Amer. Meteor. Soc., ditions govern the depth, strength, and longevity of el- 1167–1178. evated convective clouds, and can these variables be Browning, K. A., and G. B. Foote, 1976: Airflow and hail growth in supercell storms and some implications for hail suppres- observed and forecast? 4) Why do elevated supercells sion. Quart. J. Roy. Meteor. Soc., 102, 499–533. sometimes assume a linear arrangement? 5) Are el- Bryan, G. H., and J. M. Fritsch, 2001: Moist absolute instability: evated storms affected by storm outflow and surface The sixth static stability state. Bull. Amer. Meteor. Soc., 81, cold pools? If so, how? 6) Do elevated storms acquire 1207–1230. rotation in the same manner as do surface-based super- Carlson, T. N., and F. H. Ludlam, 1968: Conditions for the occurrence of severe local storms. Tellus, 20, 203–226. cells? 7) How can supercells on the cool side of baro- Colby, F. P., Jr., and B. E. Walker, 2007: Tornadoes from elevated clinic zones produce tornadoes? 8) Is there a maximum convection. Preprints, 22nd Conf. on Weather Analysis and limit to the depth of the cold air mass for elevated Forecasting and 18th Conference on Numerical Weather Pre- storms to produce tornadoes? 9) How do supercells diction, Park City, UT, Amer. Meteor. Soc., 7A.8. [Available with most unstable parcels that do not originate at the online at http://ams.confex.com/ams/pdfpapers/124653.pdf.] Colman, B. R., 1990a: Thunderstorms above frontal surfaces in surface differ from supercells that are more purely sur- environments without positive CAPE. Part I: A climatology. face based or are more purely elevated? These repre- Mon. Wea. Rev., 118, 1105–1122. sent some of the unknowns that the events discussed in ——, 1990b: Thunderstorms above frontal surfaces in environ- this paper bring to mind. As some of these questions ments without positive CAPE. Part II: Organization and in- ultimately are answered, new ones likely will be raised. stability mechanisms. Mon. Wea. Rev., 118, 1123–1144. Coniglio, M. C., H. E. Brooks, S. F. Corfidi, and S. J. Weiss, 2007: We anticipate that increased understanding of the Forecasting the maintenance of quasi-linear mesoscale con- physical processes responsible for the observed behav- vective systems. Wea. Forecasting, 22, 556–570. ior of elevated convection and castellanus will enhance Corfidi, S. F., 2003: Cold pools and MCS propagation: Forecasting the quality of both day-to-day forecasts and short-term the motion of downwind-developing MCSs. Wea. Forecast- warnings for hazardous convective weather. ing, 18, 997–1017. Davies-Jones, R. P., D. W. Burgess, and M. Foster, 1990: Test of helicity as a tornado forecast parameter. Preprints, 16th Conf. Acknowledgments. We very much thank Matthew on Severe Local Storms, Kananaskis Park, AB, Canada, Bunkers, Mike Coniglio, Robert Fovell, and Steven Amer. Meteor. Soc., 588–592. Weiss for careful, thoughtful reviews that improved the Doswell, C. A., III, 2001: Severe convective storms—An over- manuscript. We also thank our SPC colleagues for in- view. Severe Convective Storms, Meteor. Monogr., No. 50, Amer. Meteor. Soc., 1–26. formative discussions, David Bright for providing the Emanuel, K. A., 1994: Atmospheric Convection. Oxford Univer- data shown in Fig. 3d, and David Stensrud for his as- sity Press, 580 pp. sistance with Figs. 2–4. Bjorn Stevens supplied the Rain Fovell, R. G., 2005: Convective initiation ahead of the sea-breeze in Cumulus over the Ocean (RICO) images used in Fig. front. Mon. Wea. Rev., 133, 264–278. 4; otherwise, all cloud photographs were taken by the ——, G. L. Mullendore, and S. H. Kim, 2006: Discrete propagation in numerically simulated nocturnal squall lines. Mon. lead author. Partial funding for Schultz was provided by Wea. Rev., 134, 3735–3752. Vaisala, and by the NOAA/Office of Oceanic and At- Glickman, T. S., Ed., 2000: Glossary of Meteorology. 2nd ed. mospheric Research under NOAA–University of Okla- Amer. Meteor. Soc., 855 pp.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC DECEMBER 2008 FORECASTERS’ FORUM 1303

Goss, S. M., R. L. Thompson, and E. Bookbinder, 2006: An el- precipitation associated with elevated thunderstorms formed evated supercell with damaging wind from the morning of 12 in a convectively unstable layer aloft. Meteor. Appl., 5, 373– March 2006. Preprints, 23rd Conf. on Severe Local Storms, St. 384. Louis, MO, Amer. Meteor. Soc., 18.4. [Available online at ——, F. H. Glass, C. E. Graves, S. M. Rochette, and M. J. Singer, http://ams.confex.com/ams/pdfpapers/115238.pdf.] 2003: The environment of warm-season elevated thunder- Grant, B. N., 1995: Elevated cold-sector severe thunderstorms: A storms associated with heavy rainfall over the central United preliminary study. Natl. Wea. Dig., 19 (4), 25–31. States. Wea. Forecasting, 18, 861–878. Hales, J. E., Jr., 1988: Improving the watch/warning program Neiman, P. J., M. Shapiro, and L. Fedor, 1993: The life cycle of an through use of significant event data. Preprints, 15th Conf. on extratropical marine cyclone. Part II: Mesoscale structure Severe Local Storms, Baltimore, MD, Amer. Meteor. Soc., and diagnostics. Mon. Wea. Rev., 121, 2177–2199. 165–168. Rockwood, A. A., and R. A. Maddox, 1988: Mesoscale and syn- Hamblyn, R., 2001: The Invention of Clouds: How an Amateur optic scale interactions leading to intense convection: The Meteorologist Forged the Language of the Skies. Farrar Straus case of 7 June 1982. Wea. Forecasting, 3, 51–68. Giroux, 256 pp. Schmidt, J. M., and W. R. Cotton, 1989: A High Plains squall line Heymsfield, A., 1975: Cirrus uncinus generating cells and the evo- associated with severe surface winds. J. Atmos. Sci., 46, 281– lution of cirriform clouds. Part II: The structure and circula- 302. tions of the cirrus uncinus generating head. J. Atmos. Sci., 32, Scorer, R. S., 1963: Cloud nomenclature. Quart. J. Roy. Meteor. 809–819. Soc., 89, 248–253. Hobbs, P. V., and A. L. Rangno, 1985: Ice particle concentrations ——, 1972: Clouds of the World. David and Charles, 176 pp. in clouds. J. Atmos. Sci., 42, 2523–2549. Stull, R. B., 1985: A fair-weather cumulus cloud classification Horgan, K. L., D. M. Schultz, J. E. Hales Jr., S. F. Corfidi, and scheme for mixed-layer studies. J. Appl. Meteor., 24, 49–56. R. H. Johns, 2007: A five-year climatology of elevated severe ——, 1988: An Introduction to Boundary Layer Meteorology. Klu- convective storms in the United States east of the Rocky wer Academic, 670 pp. Mountains. Wea. Forecasting, 22, 1031–1044. Tardy, A., 2007: Climatology and forecasting applications for el- Houze, R. A., Jr., 1993: Cloud Dynamics. Academic Press, 573 pp. evated thunderstorms in the Great Basin and west coast of Janish, P. R., R. H. Johns, and K. C. Crawford, 1996: An evalua- the United States. Preprints, 22nd Conf. on Weather Analysis tion of the 17 August 1994 Lahoma, Oklahoma supercell/ and Forecasting. Park City, UT, Amer. Meteor. Soc. P2.12. MCS event using conventional and non-conventional analysis [Available online at http://ams.confex.com/ams/pdfpapers/ and forecasting techniques. Preprints, 18th Conf. Severe Lo- 124729.pdf.] cal Storms, San Francisco, CA, Amer. Meteor. Soc., 76–80. Thompson, R. L., C. M. Mead, and R. Edwards, 2007: Effective Ley, W. C., 1894: Cloudland: A Study on the Structure and Char- storm-relative helicity and bulk shear in supercell thunder- acteristics of Clouds. Edwards Stanford, 208 pp. storm environments. Wea. Forecasting, 22, 102–115. Ludlam, F. H., 1980: Clouds and Storms: The Behavior and Effects Trapp, R. J., D. M. Schultz, A. V. Ryzhkov, and R. L. Holle, 2001: of Water in the Atmosphere. The Pennsylvania State Univer- Multiscale structure and evolution of an Oklahoma winter sity Press, 405 pp. precipitation event. Mon. Wea. Rev., 129, 486–501. ——, and R. S. Scorer, 1957: Cloud Study: A Pictorial Guide. John Weisman, M. L., and R. Rotunno, 2000: The use of vertical wind Murray, 80 pp. shear versus helicity in interpreting supercell dynamics. J. Marwitz, J. D., 1973: Trajectories within the weak echo regions of Atmos. Sci., 57, 1452–1472. hailstorms. J. Appl. Meteor., 12, 1174–1182. World Meteorological Organization, 1956: International Cloud Moore, J. T., A. C. Czarnetzki, and P. S. Market, 1998: Heavy Atlas (Complete Atlas). Vol. 1. WMO, 175 pp.

Unauthenticated | Downloaded 09/24/21 02:08 AM UTC